Archives of the TeradataForum

Message Posted: Mon, 23 Aug 2004 @ 14:37:51 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Hello and thanks for your recommendations.

The statistics on all tables and columns in the joins are collected and up to date.

Also I'm running the query from the same machine through Queryman using the same DSN, ODBC driver version 3.02.00.04.

The only difference I see in the help session info, is that for two of the three users we have a different account string.

However it only differs in the logging part to the ampusage data ('$MML$&D&H&S' versus '$MML$&D&H').

One thing that changed recently is that I added a multi-column statistic on the table that has the user qualification on it.

The statistics are collected on a combination of article numbers and the username, which is then specified in the where clause (both articles and the username are specified in the where-part).

The user with the 'correct' explain has no rows in this table.

If the optimizer really got that smart to combine the logged-in user with the statistics it might explain the difference in the chosen plan.

However I still would like it to use the same explain for all users.

Can anybody provide some insight into what is collected for multi-column stats and how the data is used by the optimizer?

Roland

	<Prev	Next>		<<First	<Prev	Next>	Last>>